Having compressed an image,how can the compression noise or blocking artifacts be suppressed to improve the quality of the image before further processing. Thanks

  • $\begingroup$ is the "quality" with respect to the "further process" or the "final viewer" ? $\endgroup$ – Fat32 May 4 '16 at 23:33
  • $\begingroup$ Thanks. The quality is in respect to further process. My interest is not image viewing but suppressing the effect of noise for better accuracy in my final processing. $\endgroup$ – I.O Animasahun May 5 '16 at 0:44
  • $\begingroup$ ok. Most de-blocking algorithms in the block-based transform coding literature are optimized wrt human viewers. You shall somewhat adjust them to fit into the "further" processes sensitivity characteristics. $\endgroup$ – Fat32 May 5 '16 at 1:08
  • $\begingroup$ Thanks. I have tried some but with no improvement and I think adjustment is an option. I will appreciate any further suggestion that can work. I am trying to reduce the compression noise on the PRNU characteristics of an image after the image has undergone compression $\endgroup$ – I.O Animasahun May 5 '16 at 3:30
  • $\begingroup$ Which further processing are you aiming at? Depending on its properties, you may want to remove artifacts differently, because artifacts are often composite (blocking is not only a shift in the mean luminance of a block) $\endgroup$ – Laurent Duval May 5 '16 at 7:27

If you talk about JPEG-like blocking artifacts, an interesting appproach consists in the re-applications of the same JPEG algorithm on shifted versions of the image, see Enhancement of JPEG-Compressed Images by Re-application of JPEG, by Aria Nosratinia, 1999. It has a software implementation. It implement a sort of shift-invariant processing. He did the same with the JPEG-2000 coder, in Post-Processing of JPEG-2000 Images to Remove Compression Artifacts. You can find an overview of these artifacts here.

Improved methods include sparsifying priors like total variation, see for instance Adapted Total Variation for Artifact Free Decompression of JPEG Images. Recent references include Reducing Artifacts in JPEG Decompression Via a Learned Dictionary and Understanding, Optimising, and Extending Data Compression with Anisotropic Diffusion.

  • $\begingroup$ @I.O Animasahun If you give more specific details about your processing, you might get more involved techniques, but you can try the most basic ones first. $\endgroup$ – Laurent Duval May 5 '16 at 9:28
  • $\begingroup$ Thanks, I uses the Sensor pattern noise of a camera to distinguish an image coming from it or not. if the image is compressed before the extraction, this pattern is suppressed by the compression noise. So, I need a pre-processing algorithm that will reduce the artifacts due to compression so that it will not affect the overall classification accuracy. I am trying to explore your suggestion, though further suggestion is appreciated $\endgroup$ – I.O Animasahun May 6 '16 at 9:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.